Free framework
Build AI customer support that closes the hard tickets, not just the easy ones.
For support and CX leaders running regulated queues, where every reply carries rules that have to hold.
Get the framework in your inbox.

- Free
- Work email only
Why this exists
91%
of customer service leaders report executive pressure to implement AI in 2026.
Gartner, February 2026, n=321
24%
report positive financial returns from their AI investments so far.
Gartner, July 2026, n=1,303
70%
of leaders piloting agentic AI say they cannot trust and govern them.
Deloitte, August 2026, n=501 US leaders
What our framework achieves
Of the customer issues that reach the queue, our AI agents resolve 75%+ end to end, in production, at B2C operators in regulated industries across Europe.
Observed at our production deployments. Results depend on your queue.
Look inside
Three pages from the document.
Not a summary and not a brochure. This is what the framework looks like on the page.

The 8 components, one by one.

The engine that selects and enforces your rules.

The 5 questions, as a page you can cut out.
What is inside
The framework our own support agents run on, published in full.
Not a summary of it. The whole thing, so you can hold your own setup against it this week.
01
One real ticket, closed end to end
A single support ticket that needs four of your rules and four pieces of customer data, combined at the moment the reply is written. That is the actual job. It is also the point where a workflow bot and a one big prompt agent both stop.
02
The 8 components, and what enforces them
Prompts, policies, routines, glossary, macros, variables, tools and knowledge base. Your support leads write them in plain language and change them the same day. Then the part almost nobody has: how 3 of 500 rules get picked for one reply, and how an action your rules forbid is simply not available to the model.
03
The 5 questions to put to any vendor
One page you can take into your next vendor call. Run them on us first.
More resources
Read the thinking. See it shipped.

How to Build and Run AI Your Teams Actually Own
Every company is now running AI in production and almost none of them can say who owns it. Not the model. The operation: who authors the rules, who governs the behaviour, who fixes it when it is wrong.
InteractiveAI
Jul 23, 20268 min readGovernanceInteractive AgentsInsightRead
How to Choose the Infrastructure Your AI Will Run On
Once a company decides to build its own AI, it inherits an infrastructure question nobody hands it upfront. Not compute. Not models. The layer that decides whether you own what you built.
InteractiveAI
Jul 23, 202611 min readInfrastructurePlatformInsightRead
How to Manage Context for Production AI Agents
The hardest problem in production AI is not the model. It is getting the right rules and knowledge to the model at the moment of every decision, on every turn.
InteractiveAI
Jul 13, 202614 min readContextInteractive AgentsInsightRead
How to Run AI in Regulated Industries
A strategic framework for 2026 and beyond: how to put production AI on regulated systems you can govern deterministically, prove on demand, and operate under your own control, before the obligations arrive.
InteractiveAI
Jun 24, 202616 min readGovernanceTrust & SecurityInsightRead